Projects

Overview: AutoHedge is an enterprise-grade autonomous agent hedge fund that automates market analysis, risk management, and trade execution. Key features: - Multi-Agent Architecture: Specialized agents for strategy generation, technical analysis, risk assessment, and order execution. - Real-Time Market Analysis: Integration with live market data for continuous analysis and execution. - Risk-First Design: Built-in risk management and position sizing before any execution. - Structured Output: JSON-formatted recommendations and analysis for downstream systems. - Enterprise Logging: Detailed, configurable logging for audit and debugging. - Extensible Framework: Modular design for custom strategies and new venues. Tech stack: Python. Why it matters: AutoHedge is designed for institutional reliability, providing structured outputs, comprehensive logging, and a risk-first architecture that scales from single strategies to multi-venue, multi-asset deployment.

### Overview: Orpheus TTS is a state-of-the-art open-source text-to-speech system built on the Llama-3b backbone. It demonstrates the emergent capabilities of using large language models for speech synthesis. ### Key features: - **Human-Like Speech**: Natural intonation, emotion, and rhythm that surpasses state-of-the-art closed-source models. - **Zero-Shot Voice Cloning**: Allows for the creation of voices without prior fine-tuning. - **Guided Emotion and Intonation**: Control speech and emotion characteristics with simple tags. - **Low Latency**: Streamlines real-time applications with latency reduced to 200ms, further reduced to 100ms with input streaming. ### Tech stack: Python ### Why it matters: Orpheus TTS offers high-quality human-like speech with low latency, making it suitable for real-time applications and versatile across multiple languages. Its ability to clone voices and control emotions makes it a valuable tool in the field of text-to-speech technology.

Overview: This repository contains legacy code from Numenta's Hierarchical Temporal Memory (HTM) projects, focusing on anomaly detection and prediction of streaming data. The HTM theory is based on the neuroscience of the neocortex. Key features: - HTM (Hierarchical Temporal Memory) implementation - Anomaly detection - Prediction of streaming data Tech stack: Python Why it matters: This project showcases the implementation of HTM, a theory of intelligence based on the neuroscience of the neocortex, which is applicable to various machine learning and data analysis tasks.

### Overview: Decepticon is an autonomous Red Team agent designed for advanced penetration testing. It automates the engagement process, including setting up the Engagement Record of Evidence (RoE), initiating the engagement, and managing the Red Team operations. ### Key Features: - **Engagement Setup:** Automatically sets up the Engagement Record of Evidence (RoE) and initiates the engagement. - **Terminal Controls:** Offers searchable mode/model/session pickers, multiline prompts, private draft/history persistence, queued messages, independent Blue monitoring status, incident details, conversation export, and read-only `decepticon doctor`. - **Specialist Operations:** Covers reconnaissance, vulnerability research, exploitation, Active Directory, cloud, and other domains. It executes in an isolated sandbox and starts specialist workloads like BloodHound CE, Sliver C2, and others when needed. - **Autonomous Reverse Engineering (REA):** When a binary found during an engagement needs deeper analysis, the Reverser starts the optional REA workload and uses its Ghidra provider through discovered MCP tools. ### Tech Stack: - Python ### Why It Matters: Decepticon streamlines the Red Team engagement process, reducing manual effort and improving efficiency. It automates the setup and execution of complex operations, making it ideal for advanced penetration testers and security professionals looking to enhance their capabilities.

### Overview: PyTorch Metric Learning Library PyTorch Metric Learning Library is an easy-to-use tool for implementing deep metric learning models. It provides a modular, flexible, and extensible framework written in PyTorch, making it suitable for various applications. ### Key Features: - **9 Modules**: Contains 9 modules that can be used independently or combined for a comprehensive training workflow. - **Datasets Module**: Provides easy access to common datasets such as CUB200, Cars196, INaturalist 2018, and Stanford Online Products. - **Losses and Miners**: Offers a variety of loss functions and miners to enhance model performance and robustness. - **Installation Instructions**: Detailed installation guide available for easy setup. - **Documentation**: Comprehensive documentation for understanding and utilizing the library. ### Tech Stack: Python ### Why It Matters: PyTorch Metric Learning Library simplifies the process of implementing metric learning models, making it accessible to developers and researchers alike. Its modular design allows for easy integration into existing projects, and its flexibility ensures that it can be adapted to various applications without significant changes to the underlying codebase.

Overview: This repository contains the source files for the official TensorFlow documentation, including guides and tutorials. It serves as a primary resource for developers looking to learn about TensorFlow. Key features: - Source files for TensorFlow documentation - Contributions to the TensorFlow documentation through the issue tracker - Community translations for TensorFlow documentation in multiple languages Tech stack: Jupyter Notebook Why it matters: The TensorFlow documentation is essential for developers seeking to understand and utilize TensorFlow effectively. The source files provide a comprehensive resource, and community contributions ensure the content remains up-to-date and accurate.

Overview: This repository offers a collection of TensorFlow 2.x tutorials and examples, including CNN, RNN, GAN, Auto-Encoders, FasterRCNN, GPT, BERT, and more. It serves as a comprehensive resource for TensorFlow 2.0 beginners and advanced users alike. Key features: - TensorFlow 2.x tutorials and examples - CNN, RNN, GAN, Auto-Encoders, FasterRCNN, GPT, BERT, and other advanced models - TensorFlow 2.0 Stable, Beta, and Alpha releases timeline - Installation instructions for both CPU and GPU environments Tech stack: Python Why it matters: This repository provides a practical and accessible way to learn and experiment with TensorFlow 2.x, making it an invaluable resource for developers and researchers alike. Repository: dragen1860/TensorFlow-2.x-Tutorials GitHub stars: 6,342 GitHub description: TensorFlow 2.x version's Tutorials and Examples, including CNN, RNN, GAN, Auto-Encoders, FasterRCNN, GPT, BERT examples, etc. TF 2.0版入门实例代码，实战教程。

### Overview: This repository is a curated list of data science blogs, providing a collection of resources and insights from various data science enthusiasts and professionals. ### Key Features: - **RSS Feeds**: Each blog has an RSS feed available, allowing easy access to the latest content. - **Categories**: Blogs are categorized by topic, making it easy to find relevant content. - **Author Profiles**: Detailed profiles of the authors, including their blog posts and contributions. ### Tech Stack: - **Python**: The primary programming language used to create and maintain the blog posts. ### Why It Matters: This repository serves as a valuable resource for anyone interested in data science, offering a wide range of perspectives and practical tips from experienced professionals and enthusiasts. Whether you're a beginner or an advanced practitioner, this list provides a wealth of knowledge and inspiration.

### Overview: TagUI is a free, open-source Robotic Process Automation (RPA) tool designed to automate repetitive tasks on websites, desktop applications, and the command line. With its simple TagUI language, users can interact with identifiers like web elements, images, and text using OCR, making it accessible for both English and non-English users. The project is supported by open-source contributors and the community, ensuring continuous development and support. ### Key Features: - **TagUI Language**: A simple language for interacting with identifiers, including web elements, images, and text using OCR. - **Cross-Platform Compatibility**: Works on Windows, macOS, and Linux. - **MS Office Plug-ins**: Easily create and deploy Word doc as RPA robots and set up data parameters using Excel. - **Text Editor Support**: Users can create and edit robots using any text editor like Notepad, Notepad++, VS Code, Sublime, TextEdit, Vim, or the TagUI language extension for VS Code. - **Demos and Examples**: Available on GitHub, including a video demonstrating automation of data collection in multiple languages and a sample flow for interacting with a website. ### Tech Stack: JavaScript ### Why It Matters: TagUI simplifies the process of automating repetitive tasks, making it accessible to developers and non-technical users alike. Its open-source nature and community support ensure continuous development and maintenance, making it a valuable tool for anyone looking to automate their workflows.

### Overview: 《统计学习方法》（第二版）是李航教授的经典著作，涵盖了统计学习方法的各个方面，包括模型、策略和算法。本书详细介绍了分类与标注问题的解决方案，特别强调了无监督学习方法的应用。本书内容丰富，适合信息检索、自然语言处理等领域的学生和研究人员参考。 ### Key features: - **模型**：包括决策树、支持向量机、马尔可夫随机场、主成分分析、局部连接度等。 - **策略**：分类与回归方法，如感知机、k近邻法、朴素贝叶斯法。 - **算法**：线性回归、逻辑回归、随机森林、梯度下降法、马尔可夫链蒙特卡洛方法等。 - **无监督学习**：包括聚类、主成分分析、马尔可夫随机场、局部连接度等。 ### Tech stack: Python 3.5, 3.6, 3.7 ### Why it matters: 本书是统计学习方法领域的经典教材，对于理解统计学习方法的原理和应用具有重要意义。通过阅读本书，读者可以系统地掌握各种统计学习方法的理论和实践应用。 ### Repository: SmirkCao/Lihang - Primary language: Python - GitHub stars: 6325 - GitHub description: 统计学习方法, 统计学习方法(第2版)[李航] [笔记, 代码, notebook, 参考文献, Errata,

### Overview: Everywhere is an AI assistant that understands your desktop, integrating with multiple LLMs and MCP tools. It helps you act across apps seamlessly. ### Key Features: - **Context-Aware**: Recognizes and understands current app context. - **Integrated Support**: Seamlessly integrates with various apps. - **Multiple LLMs**: Utilizes multiple language models for comprehensive assistance. - **MCP Tools**: Works with Microsoft Contextual Placement for enhanced functionality. ### Tech Stack: C# ### Why It Matters: Everywhere simplifies your workflow by providing immediate, context-aware assistance across your desktop, making tasks easier and more efficient. Repository: Sylinko/Everywhere GitHub Stars: 6321

Overview: Kun is a local-first AI workspace that integrates code, documentation, and agent conversations into a single, unified runtime environment. It runs on your computer, sharing sessions, memories, and settings locally while allowing access to any compatible provider. Key features: - Code: Projects, terminals, Git, and Diff in one place, Design canvas for prototyping. - Work: Documents, papers, and materials in the same workspace. - Agent: Personal Agent, Codex, Claude Code, OpenCode, and Agent conversations with various providers. - Model: Access to any compatible provider, local model, and local gateway for configuration. Tech stack: TypeScript Why it matters: Kun simplifies collaboration by integrating code, documentation, and agent interactions into a single, local runtime, enhancing productivity and accessibility. Repository: KunAgent/Kun Primary language: TypeScript GitHub stars: 6318 GitHub description: Local-first AI agent workspace for coding, writing, design, research, and automation — one runtime for desktop GUI and TUI. README excerpt: <p align="center"> <img src="src/asset/img/kun.png" width="72" alt="Kun 蓝色 K 标识"> </p> <h1 align="center">Kun</h1> <p align="center"> 本地优先的 AI 工作台：写代码、做文档、和 Agent 一起推进事情。 </p> <p align="center"> <a href="https://github.com/KunAgent/Kun/releases">下载</a

Overview: MaiSaka, an LLM-based intelligent agent, is designed to understand and interact with users in a warm, authentic, and genuine manner, unlike traditional AI systems. Key features: - Dual language support (Chinese and English) - Web-based user interface (WebUI) - Continuous updates to improve user experience - Automatic language switching based on user preference Tech stack: Python Why it matters: MaiSaka is a significant advancement in the field of AI, offering a more human-like interaction experience that resonates with users on a deeper level.

### Overview: OpenPencil is the world's first open-source AI-native vector design tool, designed to streamline the design process by enabling designers to turn prompts into UI directly on the live canvas. It features concurrent Agent Teams, Design-as-Code, and built-in MCP Server, making it a modern alternative to traditional design tools. ### Key Features: - **Design-as-Code**: Directly turn prompts into UI on the live canvas. - **Concurrent Agent Teams**: Collaborate and work simultaneously with multiple agents. - **Built-in MCP Server**: Supports multiple models and models from different providers. ### Tech Stack: Rust ### Why It Matters: OpenPencil simplifies AI-native design by enabling designers to create and modify designs directly on the canvas, making the process more intuitive and efficient. Its ability to integrate with multiple models and provide a stable, scalable environment makes it a valuable tool for modern design workflows. Repository: ZSeven-W/openpencil GitHub Stars: 6288 GitHub Description: The world's first open-source AI-native vector design tool and the first to feature concurrent Agent Teams. Design-as-Code. Turn prompts into UI directly on the live canvas. A modern alternative to Pencil.

### Overview: This is a curated list of simple text files designed to teach AI assistants how to perform new tasks on demand, without the need for retraining. Unlike bulk-generated skill repositories, this collection focuses on real-world Agent Skills used by engineering teams, compatible with popular AI agents like Claude, Codex, and more. ### Key Features: - **Skill Directory**: A searchable directory of skills. - **Skill Installation**: Skills can be added to AI agents via CLI commands or direct URL. - **Skill Quality**: Skills are evaluated based on quality standards to ensure they are useful and effective. ### Tech Stack: TypeScript ### Why It Matters: This community-curated list provides practical, real-world skills that developers and engineers can use directly in their projects, without the need for extensive setup or retraining.

### Overview: BossSensor is a Python-based project designed to automatically hide your screen when your boss is approaching. It uses facial recognition technology to detect the presence of your boss and triggers the screen hiding mechanism. ### Key features: - Automatically hides your screen when your boss is detected. - Uses facial recognition to classify images of your boss. - Monitors changes in the environment to ensure the screen is hidden. ### Tech stack: Python ### Why it matters: BossSensor is a practical tool for employees looking to maintain privacy and avoid distractions when their boss is in the vicinity. By automating the process of hiding the screen, it streamlines the workflow and reduces the likelihood of accidental screen exposure. ### Repository: Hironsan/BossSensor ### Primary language: Python ### GitHub stars: 6290 ### GitHub description: Hide screen when boss is approaching.

### Overview: Atomic Agents is a lightweight, modular framework for building AI applications using LEGO-like components. ### Key Features: - **Single-Purpose**: Each component (agent, tool, context provider) does one thing well. - **Reusable**: Can be used in multiple pipelines. - **Composable**: Easily combines with other components. - **Predictable**: Produces consistent, reliable outputs. ### Tech Stack: Built on Instructor and Pydantic. ### Why It Matters: Atomic Agents simplifies the process of creating AI applications by adhering to software engineering principles, making it accessible and maintainable for developers. ### Repository: GitHub: https://github.com/eigenwise/atomic-agents ### GitHub Stars: 6,289 ### Primary Language: Python ### Table of Contents: - [Atomic Agents](#atomic-agents) - [What is Atomic Agents?](#what-is-atomic-agents) - [Table of Contents](#table-of-contents) - [Getting Started](#getting-started) - [Installation](#installation) - [Quick Example](#quick-example) - [Why Atomic Agents?](#why-atomic-agents) - [Core Concepts](#core-concepts) - [Anatomy of an Agent](#anatomy-of-an-agent) - [Context Providers](#context-providers) - [Chaining Schemas and Agents](#chaining-schemas-and-agents) - [Examples \& Documentation](#examples--documentation) - [Quick

### Overview: Aim is an easy-to-use and supercharged open-source experiment tracker designed to help you manage and analyze your training runs and AI metadata. ### Key Features: - **User-Friendly Interface**: Streamline your experiment tracking with a beautiful UI. - **Comparative Analysis**: Compare and observe your training runs easily. - **API Integration**: Query your data programmatically for deeper insights. ### Tech Stack: - Python ### Why It Matters: Aim simplifies the process of tracking and analyzing your experiments, making it an essential tool for developers and researchers alike.

### Overview: Tensorpack is a neural network training interface based on graph-mode TensorFlow, designed to enhance training speed and flexibility. It focuses on efficient training with minimal overhead and supports scalable data parallelism and distributed training strategies. ### Key Features: 1. **Training Speed**: Tensorpack optimizes training speed by leveraging TensorFlow's efficient execution without additional overhead. Common CNNs run on average 1.2-5x faster than equivalent Keras code. 2. **Data Loading Performance**: Tensorpack uses Python's symbolic programming to achieve high performance data loading, offering flexibility in research. 3. **Reproducible Research**: Tensorpack provides high-quality implementations of research papers, ensuring reproducibility and flexibility for actual research. 4. **Model and Data Flexibility**: Tensorpack includes only common layers, allowing the use of any TensorFlow symbolic functions, including tf.layers, Keras, Slim, and TensorLayer. ### Tech Stack: Tensorpack is built using TensorFlow and Python, focusing on speed and flexibility in neural network training. ### Why It Matters: Tensorpack is essential for researchers and developers looking to enhance the speed and flexibility of neural network training. Its focus on efficient training, scalability, and reproducibility makes it a valuable tool in the field of deep learning. ### Repository: The repository is hosted on GitHub at [tensorpack/tensorpack](https://github.com/tensorpack/tensorpack). It has received 6,285 stars, indicating its popularity and utility in the community. ### Examples: Tensorpack provides examples that demonstrate its capabilities by reproducing well-known papers.

### Overview: HyperLPR3 is a high-performance Chinese license plate recognition framework designed for developers. It offers real-time processing capabilities and supports both Android and Linux/MacOS platforms. ### Key features: - **Real-time License Plate Recognition**: Capable of processing images in real-time with a CPU performance of 720p resolution. - **Cross-Platform Compatibility**: Supports compilation and deployment on various platforms including Linux, MacOS, and Android. - **End-to-End Recognition**: Does not require character segmentation, ensuring high accuracy in both entrance and exit scenes. - **High Recognition Rate**: Achieves an accuracy of 95%-97% across various license plate types and angles. ### Tech stack: - **Python**: Used for the main application logic and integration with other libraries. - **C++**: Implemented for the core recognition algorithm and real-time processing. - **OpenCV**: Utilized for image processing and feature extraction. - **MNN**: Optimized for inference speed using the MNN library. - **OnnxRuntime**: Used for running the model on the CPU. - **FastAPI**: For the web API service. - **Uvicorn**: For the web server. - **Loguru**: For logging and debugging. - **Tqdm**: For progress bars in the CLI. - **Requests**: For HTTP requests. ### Why it matters: HyperLPR3 is a powerful tool for developers looking to integrate license plate recognition into their projects efficiently and with high performance. Its real-time capabilities and cross-platform compatibility make it suitable for a